NEW YORK: OpenAI is putting its frontier model on a cheaper footing. The company has cut developer pricing for its GPT-5.6 Sol model by more than 20 per cent for the next three months, as it faces growing competition from Anthropic and Chinese artificial intelligence models, according to Reuters.
The reductions apply to OpenAI’s application programming interface (API) and are being rolled out across eligible plans for credits used with ChatGPT Work, its agentic AI product, and Codex, its coding tool.
Consumer subscription prices are not changing. OpenAI said pricing for its Pro, Plus and Business plans remains unchanged.
Under the new pricing, GPT-5.6 Sol costs $4 per 1 million input tokens and $20 per 1 million output tokens for standard short-context use. That compares with previous prices of $5 and $30 respectively.
The move makes the model significantly cheaper for developers building applications and AI-powered services, particularly those with high volumes of model usage.
The latest reduction follows a broader pricing push by OpenAI. Last month, the company cut prices for its smaller models as businesses increasingly scrutinise their artificial intelligence spending.
OpenAI reduced pricing for its mid-tier GPT-5.6 Terra model by 20 per cent, while its lower-cost Luna model received an 80 per cent price cut.
The latest move extends that strategy to the company’s frontier offering, potentially making advanced AI capabilities more accessible to developers while putting pressure on rivals to compete on cost.
OpenAI’s price cuts come as competition intensifies across the frontier AI market.
Anthropic lists its Claude Fable 5 model at $10 per 1 million input tokens and $50 per 1 million output tokens. Its Claude Opus 5 model is priced at $5 per 1 million input tokens and $25 per 1 million output tokens.
At $4 for input and $20 for output, GPT-5.6 Sol now undercuts both Anthropic models on the listed standard pricing.
The three-month promotion signals that pricing is becoming an increasingly important battleground alongside model performance, reliability and capability. For developers, the calculation is increasingly simple: better AI is useful, but cheaper AI can scale much faster.
